Modeling power consumption of 3D MPDATA and the CG method on ARM and Intel multicore architectures

Abstract

We propose an approach to estimate the power consumption of algorithms, as a function of the frequency and number of cores, using only a very reduced set of real power measures. In addition, we also provide the formulation of a method to select the voltage-frequency scaling-concurrency throttling configurations that should be tested in order to obtain accurate estimations of the power dissipation. The power models and selection methodology are verified using two real scientific application: the stencil-based 3D MPDATA algorithm and the conjugate gradient (CG) method for sparse linear systems. MPDATA is a crucial component of the EULAG model, which is widely used in weather forecast simulations. The CG algorithm is the keystone for iterative solution of sparse symmetric positive definite linear systems via Krylov subspace methods. The reliability of the method is confirmed for a variety of ARM The researchers from Czestochowa University of Technology were supported by the National Science Centre, Poland, under Grant No. UMO-2015/17/D/ST6/04059. The researcher from Universidad Jaime I (UJI) was supported by the CICYT Project TIN2014-53495-R of MINECO and FEDER. This work was partially performed during a short-term scientific mission (STSM)

An study of the effect of process malleability in the energy efficiency on GPU-based clusters

“…With a single dimension mesh decomposition, the size of data transfers between nodes is the same, so the communication time is expected to be constant. This assumption is confirmed [20] for a group of iterative algorithms, such as MPDATA, where the halo area size is unchanged with the number of nodes greater than 1 and the size of sub-domain per node greater than the size of halo area. Figure 3 Traditionally, jobs are submitted to the system with the resource configuration that maximizes performance to complete the execution earlier (the traditional scenario).…”

“…As most stencil-based applications, MPDATA belongs to a group of memorybound algorithms. To provide high efficient memory access [20], each row of arrays is aligned to a size of 128B. For this purpose, it is required to add some extra data at the beginning of each array to align the first element of the data and an appropriate number of extra columns to provide data alignment (padding) [21] for each row.…”

CFD code adaptation to the FPGA architecture

“…To measure power consumption for the FPGA accelerator we utilize the Xilinx Board Utility and Vivado Design Suite tools. At the same time, for the CPUs, the power dissipation is measured using the Intel RAPL (Running Average Power Limit) (Rojek et al, 2017a) infrastructure.…”

“…EULAG (Rojek et al, 2017a) is an established computational model developed by the group headed by Piotr Smolarkiewicz for simulating thermo-fluid flows across a wide range of scales and physical scenarios, such as numerical weather and climate prediction, simulation of urban flows, areas of turbulence, and ocean currents, etc. EULAG is a representative of the class of anelastic hydrodynamic models.…”

Machine learning method for energy reduction by utilizing dynamic mixed precision on GPU‐based supercomputers

“…Current approaches to energy‐aware solutions (called also green computing) can be generalized as hardware‐based and software‐based . The hardware‐based ones include high‐level solutions used in HPC, tools for power measurement and energy saving, as well as low‐level solutions applied in computing kernels that can be used in traditional computer architectures (CPUs), accelerators (GPUs and Intel Xeon Phi accelerators), and mobile solutions (eg, ARM CPU, NVIDIA Tegra GPU/CPU) . Among these, PowerPack and pmlib are two frameworks to profile power and track energy consumption of serial and concurrent applications, using watt‐meters combined with software micro‐benchmarks and utilities.…”